ava i lab le at www.sc ienced i rec t . com

journa l homepage : www. e lsev ier . com/ loca te /ac toec

a c t a o e c o l o g i c a 3 2 ( 2 0 0 7 ) 2 1 5 – 2 2 3

Original article

Coexistence, habitat patterns and the assemblyof ant communities in the Yasawa islands, Fiji

Darren Ward*,1, Jacqueline Beggs

Tamaki Campus, School of Biological Sciences, University of Auckland, Private Bag 92019, Auckland, New Zealand

a r t i c l e i n f o

Article history:

Received 22 November 2006

Accepted 9 May 2007

Published online 18 June 2007

Keywords:

Assembly rules

Community structure

Co-occurrence

Fiji

Formicidae

Invasive species

a b s t r a c t

Community assembly rules are important to help understand the dynamics of biological

invasions. The coexistence of native and invasive ant species was examined by litter sam-

pling on six remote islands within the Fijian archipelago, in the Pacific Ocean. The compo-

sition of ant assemblages of the islands and also of three different habitats across islands

was very similar to each other. Estimates of species richness indicated that the sampling

effort had generally captured a large proportion of ant species (60–97%). Analysis at two dif-

ferent spatial scales (regional [islands within an archipelago], and local [plots within an is-

land]) and on two null model data sets (co-occurrence and body size), showed that the

majority (10 of 12) of assemblages were not different from randomly assembled communi-

ties. Habitat type played an important role in the co-occurrence patterns. Scrub and coco-

nut habitats, which are non-native habitats and frequently disturbed (i.e. harsh

environments), strongly influenced the assembly of the ant community. However, two in-

vasive species, Pheidole megacephala and Anoplolepis gracilipes may have also shaped the ant

communities through inter-specific competition. These two species excel at both the dis-

covery and domination of resources, and could have ‘disassembled’ the native ant fauna.

Recent surveys and ecological studies from other Pacific islands show that a very similar

set of invasive species to the Yasawa islands are ubiquitous throughout the region.

Thus, similar patterns of competition, co-occurrence and community organisation that ex-

ist in the Yasawa islands could be manifested throughout the Pacific Ocean region.

ª 2007 Elsevier Masson SAS. All rights reserved.

1. Introduction

One of the fundamental questions in community ecology is

whether assembly rules determine the order of species estab-

lishment and the structure of natural communities (Diamond,

1975; Gotelli and McCabe, 2002). Assembly rules assume that

inter-specific competition is greatest between species that

are most similar in morphology and function (and thus

resource use), and as a consequence, patterns of species co-

occurrence are manifested (MacArthur and Levins, 1967;

Diamond, 1975; Gotelli and Ellison, 2002). Gotelli and Ellison

(2002) state that competing species should co-occur less often

than expected by chance between a set of communities, and

within a community species that do co-occur should differ

substantially in body size or morphology, so that overlap

in resource utilization is reduced. Thus, there is a limit to

* Corresponding author. Tel.: þ64 9 5744223; fax: þ64 9 5744101.E-mail address: wardda@landcareresearch.co.nz (D. Ward).

1 Present address: Landcare Research, Private Bag 92170, Auckland, New Zealand.1146-609X/$ – see front matter ª 2007 Elsevier Masson SAS. All rights reserved.doi:10.1016/j.actao.2007.05.002

a c t a o e c o l o g i c a 3 2 ( 2 0 0 7 ) 2 1 5 – 2 2 3216

the similarity of species that can coexist in a community

(MacArthur and Levins, 1967; Szabo and Meszena, 2006).

Assembly rules are also important to the study of biological

invasions, in particular, whether certain rules govern the abil-

ity of invasive species to establish and spread within a native

community. Inter-specific competition and limited similarity

between species may influence the ability of new species to

invade a native community. Resident species are expected to

strongly compete with and resist the establishment of invad-

ing species that have similar resource requirements (Fargione

et al., 2003). However, the strong link between community

ecology theory and invasion biology has only recently become

apparent (Lodge, 1993; Shea and Chesson, 2002; Fargione et al.,

2003). Studies on invasive species in native communities have

the potential to examine fundamental questions of commu-

nity ecology through the interactions of species (Lodge, 1993).

Inter-specific competition is considered to be the major

structuring force of ant (Hymenoptera: Formicidae) communi-

ties (Wilson, 1971; Andersen, 1992; Morrison, 1996; Davidson,

1998; Holway, 1999; Gotelli and Ellison, 2002). Although abiotic

factors, habitat requirements, and dispersal abilities are

among several factors that can interact to shape ant com-

munities (Cole, 1983; Savolainen and Vepsalainen, 1989;

Morrison, 1996), inter-specific competition is thought to play

the major role at local spatial scales. Inter-specific competi-

tion results in dominance hierarchies being formed through

inter-specific aggression, competitive exclusion at food re-

sources and distinctive foraging strategies for either accessing

resources or avoiding dominant species (Wilson, 1971; Fellers,

1987; Savolainen and Vepsalainen, 1989; Andersen, 1992;

Davidson, 1998; Holway, 1999). In particular, dominant species

can control the spatial occurrence of other species, thus struc-

turing the ant community and creating mosaic-like patterns

of species co-occurrence (Room, 1975; Savolainen and Vepsa-

lainen, 1989). Body size could also facilitate coexistence in

tropical ant communities via differential use of habitats and

can also influence competitive interactions.

The factors shaping ant communities are numerous, and

many have been well studied, but there have been relatively

few studies examining the role of invasive ant species on com-

munity structure (Holway et al., 2002). To date, these studies

have focused on the invasion of a native community by a sin-

gle invasive ant species, principally the Argentine ant Linepi-

thema humile, or the red-imported fire ant Solenopsis invicta

(Holway et al., 2002). The interactions and coexistence be-

tween multiple invasive ant species within the context of a na-

tive ant community have seldom been examined. Morrison

(1996) examined the competitive interactions among numer-

ous invasive ant species on several remote Pacific islands,

but these islands had no native ant species. More recently,

von Aesch and Cherix (2005) have examined the native and in-

vasive ant fauna on Floreana island (Galapagos) and the com-

petitive mechanisms leading to the establishment of invasive

species. However, they did not specifically examine, or test

patterns of coexistence within and between ant assemblages.

In this paper we examine the structure of ant communities

on an island archipelago from Fiji, in the Pacific Ocean. We ex-

amine how dominance and competition affect the coexis-

tence of ant species at both local and regional scales. In

particular, we examine the influence of habitat and the

presence of invasive species on the structure of the native

ant community.

2. Methods

2.1. The Yasawa islands

Fiji lies in the central Pacific Ocean between 12 and 21� South

and between 175� West and 177� East longitudes (Evenhuis

and Bichel, 2005; Fig. 1). The Fijian archipelago consists of

over 500 islands and islets, with the two main islands, Viti

levu and Vanua levu, making up 87% of the total land area

(Smith, 1979). The Yasawa island group is approximately

40 km northwest of Viti levu (Fig. 1). The island group is

a 90 km long chain of ancient volcanic islands and consists

of 11 main islands. The Yasawa islands (and Fiji in general)

have a warm, humid tropical maritime climate with mean

monthly temperatures from 22 �C in July to 26 �C in January

(Evenhuis and Bichel, 2005). The average annual rainfall in

the Yasawa islands is approximately 1650–2290 mm, with

most of that falling in the wet season from November to April

(Evenhuis and Bichel, 2005). Overall the islands are sparsely

populated (estimated <5000 total), with permanent human

settlements on most of the larger islands. There are no roads

or motorised vehicles on the islands, and the largest-scale in-

frastructure is restricted to backpacker resorts. Almost all

movement of people and goods to and from the Yasawa is-

lands is via sea transportation from the city of Nadi on the

main island of Viti levu. Movement between islands is via

small outboard boats (<20 people).

2.2. Sampling

Sampling took place on six islands in the Yasawa group

(Kuata, Waya, Naviti, Matacawalevu, Tavewa, and Nanuya

lailai) from the 5th to 16th September 2005. Islands were cho-

sen because they represented differing degrees of size (area)

and isolation from the mainland. Two days were spent on

each island, moving south–north using a tourist catamaran

that serviced the Yasawa group daily. Local walking tracks

and topographical maps (scale 1:50,000, Fiji map series 31, edi-

tion 1 and 2, produced by the Lands and Survey Department,

Suva) were used to navigate around each island.

Three major habitats were sampled on the Yasawa islands:

deciduous coastal dry forest, scrub and coconut groves. Decid-

uous coastal dry forest is the natural forest cover on the

Yasawa islands (Watling, 2005). Canopy species include Gyro-

carpus americanus Jacq. (Hernandiaceae), Pongamia pinnata (L.)

Pierre (Fabaceae), Pleiogynium timoriense (DC.) Leenh (Anacar-

diaceae), and a common understorey species was Mallotus tilii-

folius (Blume) (Euphorbiaceae) (Smith, 1981, 1985; Watling,

2005). The second habitat (scrub) was largely a monoculture

of Leucaena leucocephala (Lam.) (Fabaceae). This is an exotic

species in Fiji, and is considered a weed on many Pacific is-

lands as it replaces indigenous vegetation. The third habitat

sampled was coconut groves, Cocos nucifera L. (Arecaceae). Co-

conuts were once a major economic crop of Fiji for the copra

industry (Smith, 1979). However, coconuts are no longer

a c t a o e c o l o g i c a 3 2 ( 2 0 0 7 ) 2 1 5 – 2 2 3 217

Fig. 1 – Fiji in relation to the western Pacific Ocean, the Yasawa island archipelago of Fiji and the six islands sampled (inset).

widely cultivated and many plantations have been

abandoned, including those on the Yasawa islands.

To sample ants a 0.5� 0.5 m quadrat was placed on the

ground and litter within the quadrat was scooped into a white

tray (30� 40� 10 cm). Litter was sifted through a 1� 1 cm wire

mesh to exclude larger debris. Sticks and rotten wood within

the quadrat were broken apart into the tray. Not all the litter

from the quadrat could be placed into the tray at one time, be-

tween 1 and 4 trays were needed. However, a standardised

time of 15 min was spent searching through the litter of

each quadrat. An aspirator was used to collect ants and trans-

fer them to a single vial of 75% ethanol. Ants were collected

from 30 quadrats on each island. Quadrats were placed hap-

hazardly on the ground and spaced at least 15 m apart, and

at least 50 m off walking trails. While on each island, ant spe-

cies were collected opportunistically by visually searching the

ground, litter, foliage, tree trunks, and inside hostels. How-

ever, these opportunistically collected species were not in-

cluded in statistical analyses. It was not possible to use

other sampling techniques such as pitfall traps or winkler

bags due to time and luggage constraints on each island.

Not all islands had the same habitat types, but habitats

were deliberately sampled in an approximate proportion to

their occurrence on each island. The islands and the number

of litter quadrats sampled from each habitat (F¼ forest,

S¼ scrub, C¼ coconut) were: Kuata (F¼ 20, S¼ 10, C¼ 0),

Waya (F¼ 23, S¼ 7, C¼ 0), Naviti (F¼ 17, S¼ 13, C¼ 0), Mataca-

walevu (F¼ 0, S¼ 10, C¼ 19), Tavewa (F¼ 10, S¼ 0, C¼ 20), and

Nanuya lailai (F¼ 10, S¼ 0, C¼ 20).

2.3. Food baiting experiment

To determine what ant species were numerically and behav-

iourally dominant, tuna baits were used to attract ants. A

grid was setup that consisted of 24 bait stations placed 5 m

apart in a 6� 4 rectangular array. At each station approxi-

mately 2 g of tuna (Sealord� chunky style tuna in spring wa-

ter) was directly placed on a white plastic index card

a c t a o e c o l o g i c a 3 2 ( 2 0 0 7 ) 2 1 5 – 2 2 3218

(7� 7 cm), on top of the leaf litter. The index card was used to

assist in the identification and counting of ants. Stations were

examined in a fixed routine, at 12, 24, 36, 48, and 60 min after

the bait was placed out. Each station was examined for 20 s.

The number of ants present on the index card of each species

was recorded, along with any behavioural interactions, de-

fined as aggression, avoidance and coexistence as described

by Human and Gordon (1999). Abundance at baits was scored

as: 1� 5 ants, 2¼ 5–9, 3¼ 10–19, 4¼ 20–50, 5� 50. Sampling

took place between 10 am and 4 pm. The identification of

most ant species could not be determined in the field. Several

specimens were collected from baits with an aspirator for

later identification, with care taken not to displace other

ants from the bait. Baiting grids were setup in four habitats:

forest (2 grids: Waya �2), scrub (3 grids: Kuata �2, Matacawa-

levu), coconut (2 grids: Matacawalevu, Tavewa), and grassland

(grass< 0.5 m tall, used for stock grazing, 5 grids: Kuata, Waya,

Naviti, Tavewa �2).

2.4. Specimen curation

Knowledge of the ant fauna of Fiji is rudimentary, and many

islands remain unexplored for ants, including the Yasawa is-

land group (Ward and Wetterer, 2006). There is no single pub-

lication to identify the ant species of Fiji. Shattuck and Barnett

(2001) was used for generic identifications, and species-level

identification was completed by examining reference speci-

mens in the New Zealand Arthropod Collection (NZAC), and

by using the following publications: Mann (1921), Wilson and

Taylor (1967), for Cardiocondyla (Seifert, 2003), Tetramorium

(Bolton, 1977, 1979), Strumigenys (Bolton, 2000; Dlussky, 1994),

Hypoponera (Wilson, 1958), and Monomorium (Heterick, 2001).

Ward and Wetterer (2006) was used to categorise species as in-

vasive, native or endemic. Taxonomic nomenclature and sub-

family classification follows Bolton (2003) and generic

classification from Bolton (1995). All specimens are held at

the NZAC.

3. Statistical analyses

3.1. Faunal composition

The number and frequency of each ant species collected was

determined for each island and habitat from litter quadrats.

Estimates of species richness and accumulation for each is-

land and habitat were made using ESTIMATES v7.0 software

(Colwell, 2005). Rarefaction curves were plotted using ob-

served species richness and the estimated number of ant spe-

cies was calculated using the Chao 2 estimator of species

richness (Colwell, 2005). The default parameters in ESTI-

MATES were used, with 50 runs. The efficiency of litter sam-

pling was evaluated using the number of observed species

divided by the Chao 2 estimate of species richness. The Shan-

non Diversity index (H0) and Simpson’s index of evenness (1/D)

were also calculated using ESTIMATES. The composition of

ant species from different habitats and islands was examined

using non-metric multidimensional scaling in PRIMER v5.0

software, using a Bray–Curtis similarity matrix on presence–

absence data from 50 runs (Clarke and Warwick, 2005).

Pairwise tests of island and habitats were examined using

Analysis of Similarities (ANOSIM) with 999 permutations.

3.2. Coexistence in litter communities

Co-occurrence of species was examined using EcoSim soft-

ware (Gotelli and Entsminger, 2005). At the regional scale,

a presence–absence matrix was constructed with each row

representing a different species, and each column represent-

ing an island. Regional analyses consisted of 3 separate matri-

ces, 1 for each habitat type (forest, scrub, and coconut), in

order to separate the effect of habitat.

At the local scale, a presence–absence matrix was con-

structed for each habitat type on each island. Each row of

the data matrix represents a different species, and each col-

umn represents a different quadrat. Thus, 12 presence–ab-

sence matrices were constructed for analysis at the local

scale, 5 matrices from forest, 4 matrices from scrub and 3 ma-

trices from coconut. The C-score was used as a metric to

quantify the pattern of co-occurrence. The observed C-score

was compared to a histogram of simulated indices from

5000 randomly constructed communities. A fixed–fixed model

setting (default) was used, where the row and column sums of

the original matrix are preserved. Thus each random commu-

nity generated by EcoSim contains the same number of spe-

cies, and the same frequency of each species as the original

community (Gotelli and Entsminger, 2005). For an assemblage

that is competitively structured, species will co-occur less

than expected (i.e. segregation), and the observed C-score

should be significantly larger than expected by chance.

At the local scale, a meta-analysis of effect sizes for co-oc-

currence patterns was used to determine the overall co-occur-

rence pattern for each habitat. The meta-analysis followed

Gotelli and Ellison (2002), where the standardized effect size

(SES) for the set of assemblages does not differ from zero.

SES is generated in EcoSim, where SES¼ (Iobs� Isim)/ssim where

Isim is the mean index of the simulated communities, ssim is

the standard deviation, and Iobs is the observed index. Com-

munities with little co-occurrence should frequently reject

the null hypothesis in the upper tail, and the meta-analysis

pattern would show an average effect size significantly greater

than zero.

Head width was used as an index of body size, a widely

used measure of size in ants (see Holldobler and Wilson,

1990). Measurements were made on 10 specimens of each spe-

cies where possible. Only the minor caste of polymorphic taxa

was used (e.g. Pheidole). Measurements were made of mounted

specimens, using an ocular micrometer calibrated with a stage

micrometer to an accuracy of 0.1 mm. Body size overlap of

coexisting species within a community was examined using

EcoSim (size overlap module) at the two spatial scales as de-

scribed above in the co-occurrence section. At the local scale,

a meta-analysis of effect sizes for variance in body size was

used to determine the overall pattern for each local site as de-

scribed in the co-occurrence section.

At both regional and local scales the EcoSim module orders

head width from the smallest to the largest measurement, cal-

culates the difference in size between two consecutive species

(segments), and from these segments a variance in segment

length (s2) is used as an index of constancy in body size ratio.

a c t a o e c o l o g i c a 3 2 ( 2 0 0 7 ) 2 1 5 – 2 2 3 219

We used the uniform body distribution option in EcoSim,

where the endpoints of the body size distribution are fixed

by the largest and smallest species in the assemblage. The

remaining species are randomly chosen from a log uniform

distribution. Observed values are compared to a null model

generated from 1000 randomly constructed communities.

The hypothesis was that a competitively structured commu-

nity should contain species that have a constant variance in

body size ratios between species compared to a randomly as-

sembled community. If coexisting species differ from one an-

other by a constant size ratio, then the segments would be

identical in length and s2¼ 0. More heterogeneity in the size

ratios of adjacent species means that the s2 will be larger. A

competitively structured community should contain species

that generate significantly smaller s2 compared to a randomly

assembled community.

The numerical and behavioural dominance of different

species were assessed using criteria from Andersen (1992)

and Davidson (1998). Numerical dominance was measured

as those species that (1) occur at a high proportion of baits;

(2) dominated baits (defined as the proportion of abundance

score of �4); and (3) that have a high average abundance score

(average of abundance scores at only those bait at which they

occurred). Interference competition was measured by (1)

aggressive behaviour (defined as the number of times a species

‘‘attacked’’ or ‘‘w as avoided’’ as a proportion of the total inter-

actions), and (2) the ability to monopolise baits (i.e. being the

only species present on baits at the end of the 60 min baiting

period). The time taken by a species to discover bait (propor-

tion of occurrence at baits at 12 min) was also examined as

a measure of exploitative competition.

4. Results

4.1. Faunal composition

Litter quadrats yielded 27 species, 17 species were native (in-

cluding six endemic to Fiji), and 10 were invasive (Table 1).

Three additional species were opportunistically collected

that were not present in the litter quadrats: two invasive spe-

cies, Tetramorium bicarinatum (Nylander) (Waya, Matacawa-

levu), and Paratrechina longicornis (Latreille) (Waya, Tavewa)

and the native Iridomyrmex anceps (Roger) (Naviti, Tavewa,

Nanuya lailai).

Three of the four species found on all six islands were inva-

sive (Table 1). A further 10 species (a mix of invasive and na-

tive) were found on four or more islands. Six species were

Table 1 – The frequency of a species collected from the Yasawa islands; from each habitat type and from all litter quadrats

Species Forest (80) Scrub (40) Coconut (59) All (179) # Islands

Endemic

Hypoponera eutrepta Wilson 0.038 0.017 2

Hypoponera monticola Mann 0.050 0.011 1

Ochetellus sororis Mann 0.050 0.011 1

Pheidole cf wilsoni Mann 0.050 0.017 0.028 3

Strumigenys chernovi Dlussky 0.013 0.006 1

Tetramorium manni Bolton 0.050 0.025 0.017 0.034 3

Native

Anochetus graeffi Mayr 0.300 0.225 0.136 0.229 5

Cardiocondyla nuda Mayr 0.063 0.175 0.153 0.117 5

Odontomachus simillimus Smith 0.213 0.220 0.168 5

Oligomyrmex atomus Emery 0.050 0.050 0.017 0.039 4

Paratrechina minutula Forel 0.438 0.025 0.085 0.229 4

Pheidole oceanica Mayr 0.013 0.100 0.028 2

Pheidole umbonata Mayr 0.150 0.068 0.089 5

Rogeria sublevinodis Emery 0.013 0.017 0.011 2

Tapinoma minutum Mayr 0.288 0.600 0.390 0.391 6

Technomyrmex albipes F. Smith 0.013 0.006 1

Tetramorium tonganum Mayr 0.275 0.075 0.140 5

Invasive

Anoplolepis gracilipes F. Smith 0.538 0.775 0.068 0.436 6

Cardiocondyla emeryi Forel 0.013 0.006 1

Monomorium fieldi Forel 0.063 0.025 0.034 0.045 4

Monomorium sechellense Emery 0.213 0.100 0.153 0.168 6

Paratrechina vaga Forel 0.363 0.186 0.223 5

Pheidole megacephala Fabricius 0.356 0.117 2

Strumigenys rogeri Emery 0.025 0.006 1

Tapinoma melanocephalum Fabricius 0.175 0.085 0.106 4

Tetramorium lanuginosum Mayr 0.038 0.050 0.028 3

Tetramorium simillimum Smith 0.250 0.275 0.169 0.229 6

Number of species 23 16 17 27

Numbers in parentheses are the number of litter quadrats. # Islands refer to the number of islands where a species was found.

a c t a o e c o l o g i c a 3 2 ( 2 0 0 7 ) 2 1 5 – 2 2 3220

only detected on one island (Table 1). In general, Chao 2 esti-

mates of species richness showed that sampling was highly

successful in capturing ant species in the litter (Table 2). Spe-

cies diversity was the lowest on the three islands that were

numerically dominated by a single species (low 1/D ratio)

(Table 2): Tavewa, which was dominated by Pheidole megace-

phala; and Naviti and Kuata, which were dominated by Anoplo-

lepis gracilipes.

Pairwise comparisons from ANOSIM showed that the over-

all ant composition of islands were very similar (Table 3). The

differences (defined as R> 0.5, Clarke and Warwick, 2005) that

existed between islands in the composition of ant species are

largely attributable to the frequency of two species, P. megace-

phala and A. gracilipes. For example, P. megacephala was wide-

spread and abundant in the coconut plantations of Tavewa,

but were absent from this habitat on Matacawalevu and

Nanuya lailai. A. gracilipes was very common in forest on Naviti

and Kuata, but was recorded only once in forest on Tavewa.

Although forest habitat had more species than other habi-

tats, sampling efficiency for forest habitat was lower than

other habitats (Table 2), indicating that other species are pres-

ent in the litter, but was not detected using the litter quadrats.

The coconut habitat had the lowest species diversity, and was

numerically dominated either by Tapinoma minutum or P. meg-

acephala. Pairwise comparisons also showed that the ant com-

position of different habitats (across all islands) was barely

separable (defined as R< 0.25, Clarke and Warwick, 2005); for-

est–scrub, R¼ 0.135; forest–coconut, R¼ 0.252; and scrub–co-

conut, R¼ 0.217.

4.2. Coexistence in litter communities

At the regional scale, observed C-scores were not significantly

different from expected C-scores generated by null models for

forest, scrub or coconut ant assemblages (forest, observed in-

dex [Iobs]¼ 0.775, mean of simulated indices [Isim]¼ 0.789,

p¼ 0.71; scrub, Iobs¼ 0.583, Isim¼ 0.565, p¼ 0.23; coconut,

Iobs¼ 0.373, Isim¼ 0.374, p¼ 0.51). These results indicate that

at the regional level assemblages were not different from

random expectation. At the local scale, ant communities

were also randomly assembled (Table 4). However, there

was some evidence (but not statistical significance) of segre-

gation for forest ant communities (as the average effect size

was greater than zero), and aggregation in scrub ant commu-

nities (the average effect size was less than zero).

Body size measurements, at the regional scale, were not

significantly different from random communities generated

by null models for scrub or coconut ant assemblages (scrub,

Iobs¼ 0.00111, Isim¼ 0.00131, p¼ 0.46; coconut, Iobs¼ 0.00564,

Isim¼ 0.00327, p¼ 0.91). These results indicate that there was

no constant ratio of body size between adjacent species. How-

ever, for forest ant assemblages, there was a greater heteroge-

neity in size ratios than expected by null models, and thus

body size in forest ant assemblages was significantly aggre-

gated (forest, observed index [Iobs]¼ 0.00445, mean of simu-

lated indices [Isim]¼ 0.00198, p¼ 0.03). At the local scale,

body size analysis mirrored the regional pattern, with forest

ant communities significantly aggregated, as the average ef-

fect size (SES) was greater than zero ( p¼ 0.004, Table 5).

Body sizes in ant communities from scrub and coconut were

randomly assembled (Table 5).

Eleven species were recorded during the baiting experi-

ment (Table 6). Overall, there was significantly more avoid-

ance behaviour at baits than attack or coexistence behaviour

(chi-square¼ 21.71, d.f.¼ 2, p< 0.001). Species that showed

high proportions of aggressive behaviour were A. gracilipes,

I. anceps and T. melanocephalum. No behavioural interactions

were observed for P. megacephala because it seldom occurred

at baits with other species. The above four species also showed

a high level of ability to monopolise baits (Table 6). There was

a positive correlation ( p¼ 0.165) between species ranked by

discovery and dominance (interference) abilities (Fig. 2).

5. Discussion

This is the first survey of ants from the Yasawa islands of Fiji,

where a total of 27 species were collected from six islands.

Table 2 – Estimates of species richness and sampling efficacy for different islands and habitats from litter quadrats

Number of Species

Island/habitat Observed Chao 2 Sampling Singletons Doubletons H0 1/D

Estimate Efficacy %

Kuata 16 18.9 84.6 4 1 2.32 8.09

Waya 19 22.2 85.5 5 2 2.58 11.49

Naviti 14 15.5 90.6 3 1 2.16 6.90

Matacawalevu 19 31.2 60.8 7 2 2.48 9.75

Tavewa 11 11.3 97.2 2 2 2.02 5.96

Nanuya lailai 13 15.9 81.8 3 0 2.37 11.43

All 27 27.9 96.7 4 2 2.75 12.77

Forest 23 32.9 70.0 5 0 2.26 12.28

Scrub 16 17.2 93.2 4 4 2.55 11.37

Coconut 17 20.9 81.3 4 1 2.18 6.53

A higher Shannon Diversity (H0) and Simpson (1/D) index means the community is more diverse.

a c t a o e c o l o g i c a 3 2 ( 2 0 0 7 ) 2 1 5 – 2 2 3 221

Estimates of species richness indicated that the sampling ef-

fort had captured a large proportion (60–97%) of the ant spe-

cies present in leaf litter. Compared to scrub and coconut

habitats, sampling was less effective in forest habitats, indi-

cating that further species remain to be detected. There

were no differences in the composition of ant assemblages be-

tween the three different habitats sampled. This was surpris-

ing, given that habitat often plays a significant role in shaping

ant communities (Morrison, 1996; Hoffmann et al., 1999).

Despite two different types of null model analyses (co-oc-

currence and body size) and analysis at two different spatial

Table 3 – Pairwise comparisons between islands (R teststatistic) of faunal composition for all habitats and withineach habitat

Pairwise comparison All Forest Scrub Coconut

Stress 0.13 0.20 0.14 0.08

Kuata, Waya 0.434 0.446 0.696

Kuata, Naviti 0.378 0.435 0.397

Kuata, Matacawalevu 0.292 0.030

Kuata, Tavewa 0.626 0.614

Kuata, Nanuya lailai 0.245 0.237

Waya, Naviti 0.292 0.372 0.044

Waya, Matacawalevu 0.219 0.383

Waya, Tavewa 0.438 0.381

Waya, Nanuya lailai 0.060 0.450

Naviti, Matacawalevu 0.064 0.141

Naviti, Tavewa 0.607 0.794

Naviti, Nanuya lailai 0.228 0.455

Matacawalevu, Tavewa 0.504 0.845

Matacawalevu, Nanuya lailai 0.150 0.299

Tavewa, Nanuya lailai 0.253 0.342 0.661

Stress (two-dimensional) is a measure of goodness-of-fit (Clarke

and Warwick, 2005). R values give an absolute measure of the sep-

aration of pairwise comparisons on a scale from �1 to 1; well sep-

arated> 0.75, clearly different> 0.5, and barely separable< 0.25

(Clarke and Warwick, 2005). R values of>0.5 are highlighted in bold.

Table 4 – Meta-analysis of effect sizes for co-occurrencepatterns at the local scale for each habitat

Habitat Lowertail

Uppertail

Averageeffectsize

SDeffectsize

t p

Forest 1 (0) 4 (2) 1.22 1.36 2.00 0.058NS

Scrub 4 (0) 0 (0) �0.78 0.52 3.01 0.028NS

Coconut 1 (0) 2 (0) 0.43 1.03 0.73 0.271NS

Numbers in the lower and upper tails indicate the number of as-

semblages for which the C-score was, respectively, less than or

greater than predicted by the null model. The number in parenthe-

ses indicates the number of assemblages with significant patterns

( p< 0.05, one-tailed test). A one-sample t-test was used to test the

hypothesis that the standardized effect size (SES) for the set of as-

semblages does not differ from zero. See Section 2 for description of

meta-analysis. Communities with little co-occurrence should fre-

quently reject the null hypothesis in the upper tail, and the meta-

analysis pattern would be an effect size significantly greater than

zero (NS, non significant; significance at p¼ 0.05/3¼ 0.016).

scales, there is little evidence to support the hypothesis that

ant assemblages in the Yasawa islands are competitively

structured. Both local and regional models generally showed

that ant communities in different habitats were randomly

assembled.

However, there is evidence to show that habitat plays an

important role in the assembly of these ant communities. At

the local scale co-occurrence patterns were considerably dif-

ferent between habitats. Local forest communities were the

only assemblages to show a segregation pattern – consistent

with assumptions regarding assembly rules and inter-specific

competition. In scrub and coconut habitats, there were ran-

dom or aggregated species co-occurrence patterns. Both scrub

and coconut habitats consist of introduced (non-native) vege-

tation to these islands. Furthermore, both these habitats are

also subject to frequent disturbance through human activities

(e.g. stock grazing, fire wood collecting), much more than the

forest habitat. Therefore, it is possible that disturbance has

acted to influence assembly rules of the ant fauna in scrub

and coconut habitats. Gotelli and Ellison (2002) also found ev-

idence that habitat type influenced the assembly of native ant

assemblages of New England, USA. They suggested that harsh

environments (habitats) were the primary filter for assembly

rules, restricting potential colonists and thus altering co-oc-

currence patterns (Gotelli and Ellison, 2002).

However, we also suggest an additional explanation for

these patterns of species co-occurrence, based on two recent

studies on invasive ants. Gotelli and Arnett (2000) and Sanders

et al. (2003) have recently shown that invasive ant species

have the ability to ‘disassemble’ native ant faunas through in-

ter-specific competition causing random or aggregated pat-

terns of species occurrence.

It is well known that invasive species often have a strong

negative impact on native ant species by exploiting similar re-

sources and by interference competition (Human and Gordon,

1999; Holway et al., 2002). Thus, we suggest that in the Yasawa

islands the native ant fauna could have been disassembled by

invasive species, primarily A. gracilipes and P. megacephala.

These two species were primarily responsible for differences

in species composition between islands (Table 3) and both

species also excel at exploiting and monopolising resources

(Table 6, Fig. 2). Additionally, the baiting experiment provides

evidence that both A. gracilipes and P. megacephala break a fun-

damental trade-off pattern. This trade-off represents an evo-

lutionary balance between exploitative and interference

Table 5 – Meta-analysis of effect sizes for body sizeoverlap patterns at the local scale

Habitat Lowertail

Uppertail

Averageeffect size

SD effectsize

t p

Forest 0 (0) 5 (0) 0.685 0.318 4.82 0.004*

Scrub 3 (0) 1 (0) �0.080 0.404 0.39 0.360NS

Coconut 1 (0) 2 (1) 0.884 1.331 1.15 0.185NS

Data organised as in Table 4. Communities with constant body size

ratios should frequently reject the null hypothesis in the lower tail,

and the meta-analysis pattern would be an effect size significantly

less than zero (NS, non significant; * significance at p¼ 0.05/

3¼ 0.016).

a c t a o e c o l o g i c a 3 2 ( 2 0 0 7 ) 2 1 5 – 2 2 3222

Table 6 – Dominance measures of species for the food baiting experiment

Species (þspecies code) Numerical dominance Interference competition Exploitative competition

Meanoccurrence

Baitsdominated

Meanabundance score

Behaviouraldominance

Monopoly Discovery

Anoplolepis gracilipes (Ag) 0.84 0.20 2.29 0.59 0.94 0.81

Pheidole megacephala (Pm) 0.96 0.95 4.82 NA 0.90 0.87

Tapinoma melanocephalum (Tme) 0.27 0.31 2.54 0.80 0.75 0.23

Iridomyrmex anceps (Ia) 0.52 0.62 3.32 0.70 0.68 0.60

Paratrechina minutula (Pmi) 0.63 0.20 2.20 0.22 0.33 0.33

Monomorium fieldi (Mf) 0.23 0.30 2.73 0.35 0.18 0.46

Tetramorium simillimum (Ts) 0.07 0.00 1.00 0.00 0.00 0.60

Paratrechina vaga (Pv) 0.28 0.11 1.96 0.26 0.00 0.56

Cardiocondyla nuda (Cn) 0.26 0.00 1.06 0.15 0.00 0.55

Tetramorium lanuginosum (Tl) 0.10 0.00 1.00 0.00 0.00 0.25

Tapinoma minutum (Tmi) 0.24 0.00 1.25 0.00 0.00 0.24

See Section 2 for definitions of each dominance measure. Species are sorted in decreasing order of monopoly. No behavioural interactions were

observed for Pheidole megacephala.

competition that promotes coexistence in ant communities

(Davidson, 1998). In the Yasawa islands there was a positive

correlation between the trade-off between dominance and

discovery. This positive correlation indicates A. gracilipes and

P. megacephala excel at both the discovery and the dominance

of resources, and thus, strongly influence the ant community

by monopolising resources through inter-specific competi-

tion. However, in native ant communities a negative correla-

tion should be evident (Davidson, 1998; Holway, 1999).

6. Conclusions

Habitat appears to act as a strong filter for the assembly of ant

communities in the Yasawa islands. Habitat type strongly

influenced patterns of species coexistence. Although we

have not fully teased apart the effects of invasive ant species

and habitat, given the results of Gotelli and Arnett (2000), and

Fig. 2 – The relationship between exploitative and

interference competition. Species are ranked by

dominance – the ability to exclude species from resources

(proportion of baits monopolised after 60 min) and

discovery – a measure of the ability to find and exploit

resources quickly (proportion of occurrence at baits at

12 min). Spearman rank correlation [ 0.165. Species codes

are given in Table 6.

Sanders et al. (2003), it seems possible that invasive ant

species in the Yasawa islands could have also disassembled

the native ant community. Recent surveys and ecological

studies from other Pacific Ocean islands show that a very sim-

ilar set of invasive species are ubiquitous throughout the re-

gion (Morrison, 1996; Wetterer, 2002; Abbott et al., 2006;

Ward and Wetterer, 2006). Thus, similar patterns of competi-

tion, co-occurrence and community organisation that exist

in the Yasawa islands could be manifested throughout the Pa-

cific Ocean region.

Acknowledgements

We thank Nathan Sanders and Margaret Stanley for com-

ments on manuscript and to Garry Barker for Fiji GIS layers.

This work was supported by Landcare Research (FRST

C09X0507), the University of Auckland, and a FRST doctoral

scholarship to DW.

r e f e r e n c e s

Abbott, K.L., Sarty, M., Lester, P.J., 2006. The ants of Tokelau. N. Z.J. Zool. 33, 157–164.

Andersen, A.N., 1992. Regulation of momentary diversity bydominant species in exceptionally rich ant communities ofthe Australian seasonal tropics. Am. Nat. 140, 401–420.

von Aesch, L., Cherix, D., 2005. Introduced ant species andmechanisms of competition on Floreana Island (Galapagos,Ecuador) (Hymenoptera: Formicidae). Sociobiology 45,463–481.

Bolton, B., 1977. The ant tribe Tetramoriini (Hymenoptera:Formicidae). The genus Tetramorium Mayr in the Oriental andIndo-Australian regions, and in Australia. Bull. Br. Mus. (Nat.Hist.) (Entomol. Ser.) 36, 67–151.

Bolton, B., 1979. The ant tribe Tetramoriini (Hymenoptera:Formicidae). The genus Tetramorium Mayr in the Malagasyregion and in the New World. Bull. Br. Mus. (Nat. Hist.)(Entomol. Ser.) 38, 129–181.

a c t a o e c o l o g i c a 3 2 ( 2 0 0 7 ) 2 1 5 – 2 2 3 223

Bolton, B., 1995. A New General Catalogue of the Ants of theWorld. Harvard University Press, Cambridge.

Bolton, B., 2000. The ant tribe Dacetini. Mem. Am. Entomol. Inst.65, 1–1028.

Bolton, B., 2003. Synopsis and classification of Formicidae. Mem.Am. Entomol. Inst. 71, 1–370.

Clarke, K.R., Warwick, R.M., 2005. Change in MarineCommunities. An Approach to Statistical Analysis andInterpretation, second ed. Plymouth Marine Laboratory,Plymouth.

Cole, B.J., 1983. Assembly of mangrove ant communities: patternsof geographic distribution. J. Anim. Ecol. 52, 339–347.

Colwell, R.K., 2005. Estimates: Statistical Estimation of SpeciesRichness and Shared Species from Samples, Version 7.5.User’s Guide and Application. Available from: http://purl.oclc.org/estimates

Davidson, D.W., 1998. Resource discovery versus resourcedomination in ants: a functional mechanism for breaking thetrade-off. Ecol. Entomol. 23, 484–490.

Diamond, J., 1975. Assembly of species communities. In: Cody, M.L., Diamond, J.D. (Eds.), Ecology and Evolution ofCommunities. Harvard University Press, pp. 342–444.

Dlussky, G.M., 1994. Ants (Hymenoptera Formicidae) of Fiji,Tonga, and Samoa and the problem of formation of islandfauna. 2. Tribe Dacetini. Entomol. Rev. 73, 110–122.

Evenhuis, N.L., Bichel, D.J., 2005. The NSF-Fiji terrestrialarthropod survey: overview. In: Evenhuis, N.L., Bickel, D.J.(Eds.), Fiji Arthropod Series I. Bishop Mus. Occas. Pap. 82, 3–25.

Fargione, J., Brown, C.S., Tilman, D., 2003. Communityassembly and invasion: an experimental test of neutral versusniche processes. Proc. Natl. Acad. Sci. U.S.A. 100,8916–8920.

Fellers, J.H., 1987. Interference and exploitation in a guild ofwoodland ants. Ecology 68, 1466–1478.

Gotelli, N.J., Arnett, A.E., 2000. Biogeographic effects of red fire antinvasion. Ecol. Lett. 3, 257–261.

Gotelli, N.J., Ellison, A.M., 2002. Assembly rules for New Englandant assemblages. Oikos 99, 591–599.

Gotelli, N.J., Entsminger, G.L., 2005. EcoSim, Null Model Softwarefor Ecologists, Version 7.72. Available from: http://www.garyentsminger.com/ecosim/index.htm

Gotelli, N.J., McCabe, D.J., 2002. Species co-occurrence: a meta-analysis of J.M. Diamond’s assembly rules. Ecology 83,2091–2096.

Heterick, B.E., 2001. Revision of the Australian ants of the genusMonomorium (Hymenoptera: Formicidae). Invertebr. Taxon. 15,353–459.

Hoffmann, B.D., Andersen, A.N., Hill, G.J.E., 1999. Impactof an introduced ant on native rain forest invertebrates:Pheidole megacephala in monsoonal Australia. Oecologia 120,595–604.

Holldobler, B., Wilson, E.O., 1990. The Ants. Harvard UniversityPress, Cambridge.

Holway, D.A., 1999. Competitive mechanisms underlying thedisplacement of native ants by the invasive Argentine ant.Ecology 80, 238–251.

Holway, D.A., Lach, L., Suarez, A.V., Tsutsui, N.D., Case, T.J., 2002.The causes and consequences of ant invasions. Annu. Rev.Ecol. Syst. 33, 181–233.

Human, K.G., Gordon, D.M., 1999. Behavioural interactions of theinvasive Argentine ant with native ant species. Insectes Soc.46, 159–163.

Lodge, D.M., 1993. Biological invasions: lessons for ecology.Trends Ecol. Evol. 8, 133–137.

MacArthur, R.H., Levins, R., 1967. The limiting similarity,convergence and divergence of coexisting species. Am. Nat.101, 377–385.

Mann, W.M., 1921. The ants of the Fiji islands. Bull. Mus. Comp.Zool. 64, 401–499.

Morrison, L.W., 1996. Community organization in a recentlyassembled fauna: the case of Polynesian ants. Oecologia 107,243–256.

Room, P.M., 1975. Relative distributions of ant species in cocoaplantations in Papua New Guinea. J. Appl. Ecol. 12, 47–61.

Sanders, N.J., Gotelli, N.J., Heller, N.E., Gordon, D.M., 2003.Community disassembly by an invasive ant species. Proc.Natl. Acad. Sci. U.S.A. 100, 2474–2477.

Savolainen, R., Vepsalainen, K., 1989. Niche differentiation of antspecies within territories of the wood ant Formica polyctena.Oikos 56, 3–16.

Seifert, B., 2003. The ant genus Cardiocondyla (Insecta:Hymenoptera: Formicidae) – a taxonomic revision of theC. elegans, C. bulgarica, C. batesii, C. nuda, C. shuckardi,C. stambuloffii, C. wroughtonii, C. emeryi, and C. minutior speciesgroups. Ann. Naturhist. Mus. Wien B Bot. Zool. 104, 203–338.

Shattuck, S.O., Barnett, N.J., 2001. Australian Ants Online.Available from: http://www.ento.csiro.au/science/ants/default.htm (accessed November 2005).

Shea, K., Chesson, P., 2002. Community ecology theory asa framework for biological invasions. Trends Ecol. Evol. 17,170–176.

Smith, A.C., 1979. Flora Vitiensis Nova. A New Flora of Fiji, vol. 1.Pacific Tropical Botanical Garden, Honolulu.

Smith, A.C., 1981. Flora Vitiensis Nova. A New Flora of Fiji, vol. 2.Pacific Tropical Botanical Garden, Honolulu.

Smith, A.C., 1985. Flora Vitiensis Nova. A New Flora of Fiji, vol. 3.Pacific Tropical Botanical Garden, Honolulu.

Szabo, P., Meszena, G., 2006. Limiting similarity revisited. Oikos112, 612–619.

Ward, D.F., Wetterer, J.K., 2006. Checklist of the ants of Fiji(Hymenoptera: Formicidae). In: Evenhuis, N.L., Bickel, D.J. (Eds.),Fiji Arthropod Series III. Bishop Mus. Occas. Pap. 85, 23–47.

Watling, D., 2005. Palms of the Fiji Islands. EnvironmentalConsultants, Suva.

Wetterer, J.K., 2002. Ants of Tonga. Pac. Sci. 56, 125–135.Wilson, E.O., 1958. Studies on the ant fauna of Melanesia. 3.

Rhytidoponera in Western Melanesia and the Moluccas. 4. Thetribe Ponerini. Bull. Mus. Comp. Zool. 119, 303–371.

Wilson, E.O., 1971. The Insect Societies. Harvard University Press,Cambridge.

Wilson, E.O., Taylor, R.W., 1967. The ants of Polynesia(Hymenoptera: Formicidae). Pac. Insects Monogr. 14, 1–109.